MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641073806 A) filed by Vellore Institute Of Technology on June 14, 2026, for Method And System For Latent Graph Densification In Sparse Social Networks For Early Rumor Detection.
Inventors include Ragavan K; Tanay Mittal; Nishant Bhatnagar; and Raj Singh.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: METHOD AND SYSTEM FOR LATENT GRAPH DENSIFICATION IN SPARSE SOCIAL NETWORKS FOR EARLY RUMOR DETECTION The present invention relates to a method and system for robust early rumor detection in sparse social interaction networks using latent semantic edge inference and graph densification. The system processes social media conversation threads by extracting textual content and user metadata associated with conversation nodes. Semantic embeddings are generated using a transformer-based model, and pairwise semantic similarity between disconnected nodes is evaluated to identify hidden semantic relationships. Latent semantic edges are automatically established between semantically related disconnected nodes and combined with explicit reply connections to construct a densified information graph configured to reduce information loss in sparse interaction structures. The densified graph is processed through a dual-stream Bidirectional Graph Attention Network (BiGAT) architecture for simultaneous propagation of textual semantic features and user metadata features. A Temporal Fusion Unit (TFU) maintains contextual continuity across temporal snapshots for improving robustness and accuracy of early-stage rumor detection in evolving social media conversations.
Disclaimer: Curated by HT Syndication.
